Going reactive with Spring WebFlux

What’s the number one rule of reactive programming with Spring WebFlux? And what happens when your team inevitably breaks it?

#1about 2 minutes

Comparing Spring MVC and Spring WebFlux models

Spring MVC uses a blocking, thread-per-request model, while Spring WebFlux uses a non-blocking, asynchronous event loop model for better resource utilization.

#2about 3 minutes

A restaurant analogy for blocking vs non-blocking I/O

An analogy of waiters in a restaurant illustrates how the non-blocking model of Spring WebFlux is more efficient than the blocking, thread-per-request model of Spring MVC.

#3about 3 minutes

How the event loop model improves resource utilization

Spring WebFlux uses a small number of event loop threads that handle requests in a non-blocking fashion, preventing threads from waiting idly on I/O operations.

#4about 4 minutes

Core concepts of Project Reactor: Publisher and Subscriber

Project Reactor is built on the Publisher-Subscriber pattern, using Mono for single items and Flux for streams, which only emit data upon subscription.

#5about 7 minutes

Building a reactive API controller with Spring WebFlux

Implementing a reactive controller involves wrapping request and response types in Mono or Flux and using a functional, chained programming style with operators like flatMap.

#6about 5 minutes

Handling blocking I/O and transactions in reactive streams

Calling blocking database operations from a reactive stream blocks the event loop, requiring the use of `subscribeOn` to offload work and breaking standard `@Transactional` behavior.

#7about 6 minutes

Implementing manual transaction management in WebFlux

Transactions in a hybrid reactive application require manual management using `PlatformTransactionManager` and a dedicated single-threaded scheduler to maintain thread context across operations.

#8about 5 minutes

Exploring reactive data access and messaging options

To fully leverage WebFlux, the data access layer must also be reactive, using technologies like R2DBC, Hibernate Reactive, or native reactive drivers for NoSQL databases.

#9about 8 minutes

Lessons from a hybrid reactive and blocking architecture

A real-world job platform demonstrates a hybrid model, highlighting the complexity of mixing blocking and non-blocking code and reinforcing that a fully reactive stack is ideal.

#10about 11 minutes

Q&A: Thread safety, RxJava vs Reactor, and debugging

The Q&A session addresses common concerns, including the importance of thread safety, the similarities between RxJava and Reactor, and the inherent difficulties of debugging asynchronous code.